1. Field of the Invention
The present invention relates generally to the field of streaming. More specifically, the present invention is related to analyzing streaming data in packetized form.
2. Discussion of Prior Art
Many electronic networks such as local area networks (LANs), metropolitan area networks (MANs), and wide area networks (WANs) are increasingly being used to transport streaming media whose real-time data transport requirements exhibit high sensitivity to data loss and delivery time distortion. The technical literature is replete with various schemes to implement Quality of Service (QOS) on such networks to address the requirements of streaming media, especially when intermixed with conventional, time-insensitive, guaranteed delivery protocol stack data traffic. Furthermore, for efficiency reasons, the streaming media transport often uses a non-guaranteed delivery upper layer protocol stack such as UDP/IP making recovery of data in the presence of packet loss difficult. Regardless of whether QOS-enabled or non-QOS-enabled networks are employed, it is necessary to monitor the behavior of packet loss, delivery time distortion, and other real-time parameters of the network to assure satisfactory quality streaming media delivery.
There exists a variety of defined Management Information Bases (MIBs) which include definitions for a number of network parameters such as packet loss, inter-arrival times, errors, percentage of network utilization, etc., whose purpose is to indicate to a network manager the general operating conditions of the network. Such traditional forms of monitoring network behavior cannot easily indicate the effects that network performance has on a single or a group of individual streaming media streams. Data gathering from MIBs operating across a range of network layers combined with a highly skilled and experienced practitioner would be required to simply determine the jitter imposed on a single MPEG video stream, for instance, and would only be possible by post-processing data gathered while the network was in operation. Determining the cause of a fault in a streaming media stream may be possible through such analysis but lacks the real-time indication of a network fault that is required to maintain high-quality networks such as for video or audio delivery. It also does not address the need to monitor large numbers of streams in real-time such as streams of Video-on-Demand (VoD) networks using less technically skilled operations personnel, as would be necessary to enable implementation of continuous cost-effective quality control procedures for widely deployed networks such as for VoD.
Histograms are often used in prior art schemes to present the arrival time behavior of packets on a network, but such histograms only represent the aggregate behavior of packets arriving at the measurement node due to the need to combine MIB data from a range of network layers to extract sufficient information to track a particular stream's performance. Traditional histograms define the jitter between any two packets. Streaming media requires more in-depth knowledge, such as the time variation across many packets referred to as the “network jitter growth”. This network jitter growth affects the streaming media quality as experienced by the user due to intermediate buffer overflow/underflow between the media source and its destination.
Network jitter growth of a media stream due to traffic congestion can also be an indicator of an impending fault condition and can thus be used to avoid transport failures rather than simply to react to faults after they occur. Conventional post-processed MIB analysis is inadequate for these purposes as described above.
The concept of regulating stream flow in a network based on the leaky bucket paradigm describes a methodology that might be used to prevent intermediate buffer overflow and packet jitter by regulating the outflow of data based on a set of parameters configured to optimize a particular flow. This does not address the need to analyze and continuously monitor multiple streams as is required during the installation and operation of networks carrying streaming media, especially for those enterprises whose revenue is derived from the high quality delivery of streaming media, such as broadcast and cable television entities.
A common prior art scheme used to effectively monitor multiple video streams is to decode each stream's MPEG content (for the video example) and display the streams on a large group of television screens. Monitoring personnel then watch the screens looking for any anomalous indications and take appropriate corrective action. This is a highly subjective and error prone process, as there is a possibility that a transient fault might be missed. This is also a reactive process, as corrective action can only be taken after a fault has occurred. Furthermore, this is also an expensive process in terms of both equipment and personnel costs. It also provides little or no indications of the root cause of the fault, thus adding to the time required for implementing corrective action. This approach also does not easily scale to modern video delivery systems based upon emerging, cost-effective high-bandwidth, networks intended to transport thousands of independent video streams simultaneously. In addition, this approach cannot pinpoint the location of the fault. To do so, the personnel and equipment must be replicated at multiple points in the distribution network, greatly increasing the cost. For this to be effective, the personnel must monitor the same stream at exactly the same time for comparison.
Many types of network delivery impairments are transient in nature affecting a limited number of packets during a period of momentary traffic congestion, for example. Such impairments or impairment patterns can be missed using traditional monitoring personnel watching video monitors. By not recognizing possible repeating impairment patterns, faults can exist for much longer periods because after the fault has passed, there is no residual trace information available for analysis. The longer a fault persists, the worse the customer satisfaction levels, and the greater the potential for lost revenues.
Whatever the precise merits, features, and advantages of the above-mentioned prior art schemes, they fail to achieve or fulfill the purposes of the present invention.